If I was actually capable of self-modification like that, and the only other part of me trying to control what I did was my consequentialist module, I would do it, But really, that would increase the willpower costs of taking/not taking actions my consequentialist module wants to take/not take anyway, and make me less moral overall as a greater fraction of my motivations became selfish ones.
If I was actually capable of self-modification like that
On our current hardware, you’re not: removing the “deontology module” isn’t really an option. However, you probably are capable of removing or changing the rules that your “deontology module” (probably not an actual module) is following. In order to feel guilty about things, you have to have experiences of guilt that have been associated to patterns of behavior, and such experiential memories are pretty open to being manipulated after the fact.
That aside, though, the reasoning you’re using in your deontology appears to be based on faulty assumptions:
sacrifice their own hedons for utilons
What evidence do you have that they’ll be sacrificing anything?
without being willing to do the same myself
This is assuming you are sufficiently similar for such an ethic to be meaningful.
In both cases, the bit you’re not taking into consideration is that trades can be non-zero-sum—a win for you and the other person both—in the case where you and your trading partner are sufficiently different in situation, values, motivations, etc. In fact, it might be unethical (in a strict utilitarian sense) for you to not make the trade, if you are sufficiently complementary to your potential trading partner.
So what you’re dealing with here isn’t so much deontology in a true moral sense, it’s just cached guilt based on a simplistic idea of fairness (and perhaps a cached assumption that if you’re getting something out of it, it’s probably hurting someone and you should feel bad).
make me less moral overall as a greater fraction of my motivations became selfish ones.
Yeah… that part really raises my probability estimate for a cached guilt-tripping pattern. One thing I’ve noticed in a lot of people is that they’re actually raised in such a way as to end up believing that they don’t have any motivations of their own that aren’t “bad”, and then believe they need rules to make themselves behave properly.
The flaw in this logic is that if you aren’t a good person, then why are you motivated to follow the rules?
This pattern is extremely common in self-defeating behaviors: we get taught our intrinsic motivations are bad, and that so we must follow some set of rules in order to have any chance of being a good person. Following these (usually extreme) rules, however, doesn’t actually get all our basic human needs met, in part because we’ve learned those needs are evidence of our evilness or selfishness, so we don’t even think of them as “needs”!
Then, because the needs don’t get met, we experience strong and increasing intrinsic motivation to meet them in whatever way possible, until self-control fails… which then seems to confirm the original belief that our essential nature is bad and needs to be controlled.
But if the needs were recognized as valid and part of the same entity that desires to be good in the first place, none of this would be a problem: you could have good stuff for you and other people, now and in the future. (That is, your notion of utility would include terms for other people and for your future selves, not just the present-moment one who would like to indulge in something harmful to others or your future.)
In other words, our desire to be a “good person” (whatever definition we have for that) is just as much a part of our consequentialist brains’ motivation as our desire for short-term treats at the expense of others or our long-term goals. Forcing all your conscious decisions to pass through admonitions to “be good” just proves this point, because if you didn’t want to be good, there’d be no reason to admonish yourself.
That is, actually using this strategy proves that there’s an error in a key factual assumption: that you don’t intrinsically want to be a good person and desire to good things of your own, unrestrained free will.
Come to think of it, this is actually a bit like Friendly AI, in the sense that this particular kind of self-admonishment is like thinking yourself an UnFriendly AI that thus needs to be kept under restraint in order not to do something bad. However, if you were actually unfriendly, you would have no desire to restrain yourself!
OTOH, if you were Friendly, you might choose to voluntarily limit your options or precommit to self-restraint, on the assumption that you might contain bugs you can’t detect, or if you are aware of bugs that cause friendliness-failure under certain conditions.
In the human case, that’d be like using rules to patch specific cases where hardware flaws result in underweighting certain factors at decision-making time. The catch is that while this is extremely useful for managing children in non-ancestral environments (e.g. schools), it’s mostly not a functional strategy for a full-grown adult—in fact it’s a complete waste of one’s self-control circuitry to use it as a primary decision-making strategy!
The desire to do things for my own enjoyment. Usually they are short-sighted too. Such as “Forget the work from home you could be doing right now, which could be making money for efficient altruism. Instead, look at pictures of cats on the internet.”
The problem (from the perspective of the altruistic part of me) is that I’m trading off a lot of well-being of strangers for a little of my own well-being.
If I was actually capable of self-modification like that, and the only other part of me trying to control what I did was my consequentialist module, I would do it, But really, that would increase the willpower costs of taking/not taking actions my consequentialist module wants to take/not take anyway, and make me less moral overall as a greater fraction of my motivations became selfish ones.
On our current hardware, you’re not: removing the “deontology module” isn’t really an option. However, you probably are capable of removing or changing the rules that your “deontology module” (probably not an actual module) is following. In order to feel guilty about things, you have to have experiences of guilt that have been associated to patterns of behavior, and such experiential memories are pretty open to being manipulated after the fact.
That aside, though, the reasoning you’re using in your deontology appears to be based on faulty assumptions:
What evidence do you have that they’ll be sacrificing anything?
This is assuming you are sufficiently similar for such an ethic to be meaningful.
In both cases, the bit you’re not taking into consideration is that trades can be non-zero-sum—a win for you and the other person both—in the case where you and your trading partner are sufficiently different in situation, values, motivations, etc. In fact, it might be unethical (in a strict utilitarian sense) for you to not make the trade, if you are sufficiently complementary to your potential trading partner.
So what you’re dealing with here isn’t so much deontology in a true moral sense, it’s just cached guilt based on a simplistic idea of fairness (and perhaps a cached assumption that if you’re getting something out of it, it’s probably hurting someone and you should feel bad).
Yeah… that part really raises my probability estimate for a cached guilt-tripping pattern. One thing I’ve noticed in a lot of people is that they’re actually raised in such a way as to end up believing that they don’t have any motivations of their own that aren’t “bad”, and then believe they need rules to make themselves behave properly.
The flaw in this logic is that if you aren’t a good person, then why are you motivated to follow the rules?
This pattern is extremely common in self-defeating behaviors: we get taught our intrinsic motivations are bad, and that so we must follow some set of rules in order to have any chance of being a good person. Following these (usually extreme) rules, however, doesn’t actually get all our basic human needs met, in part because we’ve learned those needs are evidence of our evilness or selfishness, so we don’t even think of them as “needs”!
Then, because the needs don’t get met, we experience strong and increasing intrinsic motivation to meet them in whatever way possible, until self-control fails… which then seems to confirm the original belief that our essential nature is bad and needs to be controlled.
But if the needs were recognized as valid and part of the same entity that desires to be good in the first place, none of this would be a problem: you could have good stuff for you and other people, now and in the future. (That is, your notion of utility would include terms for other people and for your future selves, not just the present-moment one who would like to indulge in something harmful to others or your future.)
In other words, our desire to be a “good person” (whatever definition we have for that) is just as much a part of our consequentialist brains’ motivation as our desire for short-term treats at the expense of others or our long-term goals. Forcing all your conscious decisions to pass through admonitions to “be good” just proves this point, because if you didn’t want to be good, there’d be no reason to admonish yourself.
That is, actually using this strategy proves that there’s an error in a key factual assumption: that you don’t intrinsically want to be a good person and desire to good things of your own, unrestrained free will.
Come to think of it, this is actually a bit like Friendly AI, in the sense that this particular kind of self-admonishment is like thinking yourself an UnFriendly AI that thus needs to be kept under restraint in order not to do something bad. However, if you were actually unfriendly, you would have no desire to restrain yourself!
OTOH, if you were Friendly, you might choose to voluntarily limit your options or precommit to self-restraint, on the assumption that you might contain bugs you can’t detect, or if you are aware of bugs that cause friendliness-failure under certain conditions.
In the human case, that’d be like using rules to patch specific cases where hardware flaws result in underweighting certain factors at decision-making time. The catch is that while this is extremely useful for managing children in non-ancestral environments (e.g. schools), it’s mostly not a functional strategy for a full-grown adult—in fact it’s a complete waste of one’s self-control circuitry to use it as a primary decision-making strategy!
What exactly are selfish motivations?
The desire to do things for my own enjoyment. Usually they are short-sighted too. Such as “Forget the work from home you could be doing right now, which could be making money for efficient altruism. Instead, look at pictures of cats on the internet.”
What’s wrong with doing things for your own enjoyment if you value them more highly than the well-being of strangers?
The problem (from the perspective of the altruistic part of me) is that I’m trading off a lot of well-being of strangers for a little of my own well-being.
Isn’t the well-being of strangers a component of your well-being? (Assuming you care about them.)